Aunque dudo que haya una forma cerrada, podemos obtener una integral doble con bastante facilidad:
$$\frac{1}{\sqrt{n}}=\frac{2}{\sqrt{\pi}} \int_0^\infty e^{-n x^2} dx$$
$$\frac{1}{n+1}= \int_0^\infty e^{-(n+1) y} dy$$
Then the general term will be:
$$\frac{1}{\sqrt{n}(n+1)}=\frac{2}{\sqrt{\pi}} \int_0^\infty \int_0^\infty e^{-y} e^{-n (x^2+y)} ~dx ~dy$$
Let's try exchanging the summation and integration:
$$\sum_{n=1}^\infty e^{-n (x^2+y)}=\frac{1}{e^{x^2+y}-1}$$
Now we need the integral to converge, otherwise this whole thing would be in vain.
Numerically, using Mathematica, we have:
$$\frac{2}{\sqrt{\pi}} \int_0^\infty \int_0^\infty \frac{e^{-y}}{e^{x^2+y}-1} ~dx ~dy=1.8600250788 \dots$$
Which is quite in good agreement with the numerical value obtained for the series itself (the OEIS gives $1.86002507922119\puntos de dólares).
Así, podemos afirmar:
$$\sum_{n=1}^\infty\frac{1}{\sqrt n(n+1)}=\frac{2}{\sqrt{\pi}} \int_0^\infty \int_0^\infty \frac{e^{-y}}{e^{x^2+y}-1} ~dx ~dy$$
Por qué no puedo tratar de obtener una sola integral? Porque en ese caso después de la sumatoria bajo el signo integral tengo sólo divergentes de las integrales.
Editar:
A partir de la OEIS vínculo de hecho podemos ver que el único integral también funciona. Eso es lo que he intentado (solo usando mi primera integral y resumiendo el resultado de la con $1/(n+1)$, pero pensé que el resultado de la integral de $$\int_0^\infty (-1-e^{x^2} \log (1-e^{-x^2}))dx$$ diverged. Apparently it doesn't.
Edit 2:
Let's try to make my double integral a little prettier.
$$\int_0^\infty \int_0^\infty \frac{e^{-y}}{e^{x^2+y}-1} ~dx ~dy=2\int_0^\infty \int_0^\infty \frac{z~e^{-z^2}}{e^{x^2+z^2}-1} ~dx ~dz=$$
$$x=r \cos t \\ z= r \sin t$$
$$=2\int_0^\infty \int_0^{\pi/2} \frac{r^2 \sin t~e^{-r^2 \sin^2 t}}{e^{r^2}-1} ~dt ~dr= \\ =2\int_0^\infty \int_0^1 \frac{r^2 e^{-r^2} e^{r^2 u^2}}{e^{r^2}-1} ~du ~dr=\sqrt{\pi} \int_0^\infty \frac{r ~\text{erfi}(r) e^{-r^2}}{e^{r^2}-1} ~dr$$
Esta integral converge muy bien y obtenemos:
$$\sum_{n=1}^\infty\frac{1}{\sqrt n(n+1)}=2 \int_0^\infty \frac{r ~\text{erfi}(r) e^{-r^2}}{e^{r^2}-1} ~dr$$
Mathematica fácilmente da 50 correcto de dígitos de la serie:
In[71]:= NIntegrate[
2 u Erfi[u] Exp[-u^2]/(Exp[u^2] - 1), {u, 0, Infinity},
WorkingPrecision -> 50]
Out[71]= 1.8600250792211903071806959157171433246665241215235
O podemos reescribir esto en un poco más corto formulario:
$$\sum_{n=1}^\infty\frac{1}{\sqrt n(n+1)}=\int_0^\infty \frac{\text{erfi}(\sqrt{r}) e^{-r}}{e^{r}-1} ~dr$$
We can also generalize the above to:
$$\sum_{n=1}^\infty\frac{1}{\sqrt n(n+a)}=\frac{1}{\sqrt{a}}\int_0^\infty \frac{\text{erfi}(\sqrt{ar}) e^{-ar}}{e^{r}-1} ~dr$$
Taking the limit for $un \a 0$, we recover the well known integral for the zeta function $\zeta(3/2)$.
Edit 3:
I'm sorry for bloating this post, but I wanted to make a few notes about numerical integration of the obtained expression.
It can be easily done by Gauss-Laguerre quadrature, provided we remove the singularity near $r=0$. Which can be done by getting the asymptotic expression as $r \a 0$:
$$\frac{\text{erfi}(\sqrt{r})}{e^{r}-1} \to \frac{2}{\sqrt{\pi r}}-\frac{\sqrt{r}}{3\sqrt{\pi}}+\frac{r^{3/2}}{30\sqrt{\pi}}+\frac{r^{5/2}}{315\sqrt{\pi}}+\dots$$
This singularity is exactly integrable and so we can write:
$$\int_0^\infty \frac{\text{erfi}(\sqrt{r}) }{e^{r}-1} e^{-r}~dr=2-\int_0^\infty
\left(\frac{2}{\sqrt{\pi r}}-\frac{\text{erfi}(\sqrt{r}) }{e^{r}-1} \right) e^{-r}~dr$$
This allows us to approximate the integrand near $0$ by a well defined series:
$$\frac{2}{\sqrt{r}}-\sqrt{\pi} \frac{\text{erfi}(\sqrt{r})}{e^{r}-1}= \\ =\sqrt{r} \left(3-\frac{r}{30}-\frac{r^2}{315} +\frac{r^3}{1512}+\frac{r^4}{13860}-\frac{521 r^5}{32432400}-\frac{29 r^6}{16216200}+\dots \right) $$
The terms shown give the absolute error of the order $10^{-11}$ at $r=1/4$.
So we can denote:
$$f(r)=\frac{\sqrt{r}}{\sqrt{\pi}} \left(3-\frac{r}{30}-\frac{r^2}{315} +\frac{r^3}{1512}+\frac{r^4}{13860}-\frac{521 r^5}{32432400}-\frac{29 r^6}{16216200} \right) $$
And write:
$$\sum_{n=1}^\infty\frac{1}{\sqrt n(n+1)} \approx 2-\int_0^{1/4}
f(r) e^{-r}~dr-\int_{1/4}^\infty
\left(\frac{2}{\sqrt{\pi r}}-\frac{\text{erfi}(\sqrt{r}) }{e^{r}-1} \right) e^{-r}~dr$$
The first integral can taken exactly in terms of error function, while the second can be approximated by Gauss-Laguerre quadrature with very good accuracy.
Provided both the integrals above are computed with high enough precision, we get absolute error of the order $10^{-12}$ cuando se compara con el valor numérico de la serie.
Esto también está de acuerdo con el obligado de @robjohn la respuesta.
No es muy conveniente, pero al menos me han demostrado que el uso de esta integral si no se tiene el software avanzado como Mathematica.